Bank response to higher capital requirements - EconStor

econstor Make Your Publications Visible. A Service of zbw Leibniz-Informationszentrum Wirtschaft Leibniz Information Centre for Economics Gropp, Reint; Mosk, Thomas; Ongena, Steven; Wix, Carlo Working Paper Bank response to higher capital requirements: Evidence from a quasi-natural experiment IWH Discussion Papers, No. 33/2016 Provided in Cooperation with: Halle Institute for Economic Research (IWH) – Member of the Leibniz Association Suggested Citation: Gropp, Reint; Mosk, Thomas; Ongena, Steven; Wix, Carlo (2016) : Bank response to higher capital requirements: Evidence from a quasi-natural experiment, IWH Discussion Papers, No.

33/2016, Leibniz-Institut für Wirtschaftsforschung Halle (IWH), Halle (Saale) This Version is available at: http://hdl.handle.net/10419/148361 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen.

Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public.

If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence.

www.econstor.eu

Discussion Papers No.33 December 2016 Bank Response To Higher Capital Requirements: Evidence From A Quasi-natural Experiment Reint E. Gropp, Thomas Mosk, Steven Ongena, Carlo Wix

II IWH Discussion Papers No. 33/2016 Authors Reint E. Gropp Halle Institute for Economic Research (IWH) – Member of the Leibniz Association, and Otto-von-Guericke-University Magdeburg E-mail: reint.gropp@iwh-halle.de Tel +49 345 7753 700 Thomas Mosk Goethe University Frankfurt and Sustainable Architecture for Finance in Europe (SAFE) E-mail: mosk@safe.uni-frankfurt.de Steven Ongena University of Zurich, Swiss Finance Institute, KU Leuven and CEPR E-mail: steven.ongena@bf.uzh.ch Carlo Wix Goethe University Frankfurt and Sustainable Architecture for Finance in Europe (SAFE) E-mail: wix@safe.uni-frankfurt.de The responsibility for discussion papers lies solely with the individual authors.

The views expressed herein do not necessarily represent those of the IWH. The papers represent preliminary work and are circulated to encourage discussion with the authors. Citation of the discussion papers should account for their provisional character; a revised version may be available directly from the authors. Comments and suggestions on the methods and results presented are welcome. IWH Discussion Papers are indexed in RePEc-EconPapers and in ECONIS. Editor Halle Institute for Economic Research (IWH) – Member of the Leibniz Association Address: Kleine Maerkerstrasse 8 D-06108 Halle (Saale), Germany Postal Address: P.O.

Box 11 03 61 D-06017 Halle (Saale), Germany Tel +49 345 7753 60 Fax +49 345 7753 820 www.iwh-halle.de ISSN 2194-2188

III IWH Discussion Papers No. 33/2016 We study the impact of higher capital requirements on banks’ balance sheets and its transmission to the real economy. The 2011 EBA capital exercise provides an almost ideal quasi-natural experiment, which allows us to identify the effect of higher capital requirements using a difference-in-differences matching estimator. We find that treated banks increase their capital ratios not by raising their levels of equity, but by reducing their credit supply. We also show that this reduction in credit supply results in lower firm-, investment-, and sales growth for firms which obtain a larger share of their bank credit from the treated banks.

Keywords: banking, regulation, real effects of finance JEL Classification: E22, E44, G21 Bank Response To Higher Capital Requirements: Evidence From A Quasi-natural Experiment* Abstract * We thank our discussants Jean Helwege and Asaf Manela and conference participants at the 2016 Western Finance Association Meeting and the FDIC/JFSR 16th Annual Bank Research Conference. Furthermore, we appreciate helpful comments from Ralph de Haas, Rainer Haselmann, Sascha Steffen and seminar participants at Goethe University Frankfurt, the European Bank for Reconstruction and Development, and the Halle Institute for Economic Research – Member of the Leibniz Association.

Wix gratefully acknowledges financial support from the Research Center SAFE, funded by the State of Hessen research initiative LOEWE, and from the Halle Institute of Economic Research – Member of the Leibniz Association.

Basel III, which will become fully effective in 2019, significantly increases capital requirements for banks. However, at this point the economic implications of such higher capital requirements are still unclear. Banks can, in principle, increase their regulatory capital ratios in two different ways: they can either increase their levels of regulatory capital (the numerator of the capital ratio) or they can shrink their risk-weighted assets (the denominator of the capital ratio) (Admati, DeMarzo, Hellwig, and Pfleiderer, 2010). While raising capital is generally considered “good deleveraging” by regulators, shrinking assets has potentially adverse effects if many banks simultaneously engage in cutting lending (Hanson, Kashyap, and Stein, 2011).

How banks adjust their balance sheets in response to higher capital requirements is thus an empirical question of crucial importance for understanding the real implications of the higher capital requirements recently imposed under Basel III.

The empirical identification of the effect of higher capital requirements on banks’ behavior faces a number of challenges. The most important challenge is to find exogenous variation in capital requirements. Yet, capital requirements tend to vary little over time, and when they do change, they change for all banks in a given economic area at the same time, leaving no crosssectional variation to exploit. In the case when supervisors make use of discretion and impose bank-specific requirements, they will be correlated with (unobserved) bank characteristics and thus not be exogenous with regard to banks’ balance sheets.

Finally, in order to assess the effects of capital requirements on bank lending, one needs to disentangle credit supply from credit demand. We address these empirical challenges by exploiting the 2011 capital exercise, conducted by the European Banking Authority (EBA), as a quasi-natural experiment. The capital exercise required a subset of European banks to reach and maintain a 9% core tier 1 capital ratio by the end of June 2012.1 The institutional features of the capital exercise are particularly well-suited to address the above mentioned empirical challenges. First, the required core tier 1 ratio of 9% constituted an economically significant increase in capital requirements compared to the previously required 5%.2 1 The core tier 1 ratio is defined as a bank’s core tier 1 capital over a bank’s risk-weighted assets, with core tier 1 capital comprising only the highest quality capital instruments (common equity), disclosed reserves and hybrid instruments provided by governments (EBA, 2011b) 2 Two regulatory interventions by the EBA increased the capital requirements for EBA banks in 2011: the 2011 EBA stress test required a 5% core tier 1 ratio, and the 2011 EBA capital exercise subsequently raised the required core tier 1 ratio to 9%.

The estimated 115 billion euro capital shortfall due to the EBA capital exercise was however well above the 2.5 billion euro capital shortfall due to the 2011 EBA stress test (Acharya, Engle, and DianePierret, 2014). As we argue in Section 2, we therefore focus on the 2011 EBA capital exercise as the main shock to EBA banks’ capital requirements.

Second, and more importantly, the rule by which banks were selected to participate in the capital exercise allows us to disentangle the effect of capital requirements from effects associated with bank size. The EBA used a country-specific selection rule and included banks “in descending order of their market shares by total assets in each Member State” such that the exercise covered “50% of the national banking sectors in each EU Member State” (EBA, 2011a).3 Since national banking sectors in Europe differ with regard to their total size, this country-specific selection threshold yielded a considerable overlap in size between banks participating and not participating in the exercise.

Moreover, the explicit selection rule based on bank size implies that selection into the capital exercise was based on observable bank characteristics. We exploit this exogenous variation in the bank selection rule and employ a difference-in-differences matching estimation approach to examine how banks subject to higher capital requirements adjust their balance sheets compared to otherwise similar banks not subject to a change in capital requirements. Our main findings are as follows. First, we document that EBA banks raised their core tier 1 capital ratios by 1.9 percentage points compared to banks not subject to the higher capital requirements.4 EBA banks achieved this by reducing their levels of risk-weighted assets by 16 percentage points rather than by increasing their levels of capital relative to the matched control group.

The control group is crucial for uncovering this finding: EBA banks increased their levels of core tier 1 capital by 21% over our sample period, but similar large European banks in the control group raised their levels of core tier 1 capital by the same magnitude. Banks can, in principle, reduce risk-weighted assets in two different ways: they can either shift from riskier assets into safer assets while keeping total asset size constant (risk reduction), or they can reduce total asset size while keeping the average asset risk constant (asset shrinking). We find that EBA banks reduced their risk-weighted assets relative to the matched control group by engaging in asset shrinking rather than risk reduction.

We show that this reduction in total assets can mainly be attributed to a reduction in outstanding customer loans.

Simply observing a reduction in outstanding customer loans on banks’ balance sheets is, however, not sufficient to conclude that the supply of credit by EBA banks contracted, since this might 3 The EBA used the same selection procedure as in the EBA stress test in June 2011. Selection was based on total consolidated assets as of end of 2010 and therefore not based on bank-specific events in the months prior to the capital exercise. 4 We adopt the following terminology: “EBA Banks” are banks participating in the 2011 EBA capital exercise; “Non-EBA banks” are other European banks not participating in the 2011 EBA capital exercise.

very well just reflect a reduction in credit demand by firms borrowing from EBA banks. In order to disentangle credit supply from credit demand, we use syndicated loan data and exploit the presence of multiple bank-firm relationships to control for credit demand. Specifically, we employ a modified version of the Khwaja and Mian (2008) estimator, which estimates the change in outstanding syndicated loans of a bank to country-industry firm clusters (see Acharya, Eisert, Eufinger, and Hirsch (2016); De Jonghe, Degryse, Jakovljevic, Mulier, and Schepens (2016a); De Jonghe, Dewachter, Mulier, Ongena, and Schepens (2016b)).

In the loan-level part of the paper, we show that EBA banks reduced their credit supply of syndicated loans by 27 percentage points relative to banks in the control group.

Ultimately, the degree to which a reduction in credit supply from EBA banks implies real effects at the firm level depends on the extent to which other banks, not subject to higher capital requirements, “pick up the slack”. Hence, we investigate how EBA banks’ reduction in lending due to the increase in capital requirements affects the growth of firms which obtain a larger share of their bank credit from EBA banks. We find that firms with a high EBA borrowing share exhibited 4 percentage points less asset growth, 6 percentage points less investment growth, and 5 percentage points less sales growth than firms less reliant on funding from EBA banks.

This result is driven by unlisted firms which are less likely to substitute a reduction in credit supply with other sources of funding.

A number of placebo and falsification tests suggest that our results are not confounded by other factors. To rule out that our results are driven by EBA banks’ exposure to the European sovereign debt crisis, we conduct a placebo test around the start of the crisis in 2010 and show that EBA banks and banks in the matched control group exhibited a similar evolution in their levels of core tier 1 capital and risk-weighted assets during this placebo period.5 Other contaminating events, such as moral suasion by governments or the ECB’s longer-term refinancing operations (LTRO), could provide alternative explanations for our results.

Predominately domestic banks in Greece, Ireland, Italy, Portugal, and Spain (GIIPS countries) increased their exposures to domestic sovereign debt (see Becker and Ivashina (2014); Ongena, Popov, and Van Horen (2016)) and made use of the ECB’s LTRO program (Van Rixtel and Gasperini, 2013). We therefore test whether our 5 Popov and Van Horen (2015) show that banks with sovereign exposures already reduced their lending in 2010, one year prior to the capital exercise.

results are driven by banks from these countries, but we do not find evidence for this alternative hypothesis. To further rule out that our results are driven by unobserved country-year-specific factors, we additionally conduct a difference-in-differences regression analysis including countryyear fixed effects and show that this does not affect our main results. Our paper is most closely related to the literature examining the effect of shocks to banks’ capital on bank lending. Peek and Rosengren (1997) exploit an exogenous shock to bank capital without a change in capital requirements to indirectly infer the effect on lending when capital requirements become binding.

Another strand of literature seeks to directly exploit changes in capital requirements. An early study by Berger and Udell (1994) investigates bank lending before and after the introduction of Basel II, but without the benefit of exogenous cross-sectional variation in capital requirements. To alleviate this concern, Kashyap, Stein, and Hanson (2010) adopt a model-based calibration approach for the U.S., Fraisse, Lé, and Thesmar (2015) exploit variation in capital requirements across banks in France due to the use of internal risk models, Aiyar, Calomiris, Hooley, Korniyenko, and Wieladek (2014) study the impact of changes to U.K.

bank-specific capital requirements on cross-border bank loan supply, Jimenéz, Ongena, Peydró, and Saurina (2016) analyze the introduction and later modifications in dynamic provisioning requirements in Spain, and Kisin and Manela (2016) estimate the shadow cost of capital requirements using data on a costly loophole that allowed banks in the U.S. to relax these constraints. More recently, Célérier, Kick, and Ongena (2016) explore the impact on lending in Germany by banks affected by tax reforms in Italy (in 2000) and Belgium (in 2006) which decreased their cost of bank equity. Hence, these papers show in various single-country settings that increasing capital requirements (or cost) leads banks to contract lending, though only moderately so in good times.

Our paper contributes to this literature in a number of ways. First, we exploit the countryspecific bank selection rule of the 2011 EBA capital exercise to identify the effects of higher capital requirements across 21 countries during times of economic distress (Acharya, Engle, and DianePierret, 2014).6 Second, we do not only focus on bank lending, but investigate in detail how banks choose to adjust both the assetand liability side of their balance sheets to comply with higher capital requirements. We provide empirical evidence for the recent theoretical prediction by Admati, 6 Mésonnier and Monks (2015) study the effect of the EBA capital exercise on bank lending, but do not exploit the selection rule and do not use loan-level data to disentangle credit supply from credit demand.

DeMarzo, Hellwig, and Pfleiderer (2016) that banks’ existing shareholders prefer to increase their capital ratios by reducing risk-weighted assets instead of raising new capital. Third, we study the transmission of banks’ balance sheet adjustments to the firm level and assess the effect of higher capital requirements on the real economy. I. The 2011 EBA Capital Exercise This section describes the objective and institutional details of the EBA capital exercise, which was announced by the EBA on October 26, 2011 (see Figure 1). The objective of the exercise was to restore confidence in the EU banking sector by ensuring that banks had sufficient capital to insure against unexpected losses.

To achieve this objective, the EBA required 61 banks to build up additional capital buffers to reach a level of 9% core tier 1 ratio by the end of June 2012.7 [Figure 1 about here] The capital exercise came mostly unexpected only a few months after the 2011 EBA stress test in June. For example, the Financial Times (2011) reported after the first announcement of the capital exercise on October 11, 2011 that the 9% requirement was “well beyond the current expectations of banks and analysts.” The credibility and rigor of the June stress test had been criticized, in particular because the Belgian bank Dexia was declared in the stress test to be one of the safest banks in Europe, but had failed less than three months later (Greenlaw, Kashyap, Schoenholtz, and Shin, 2012).

Although both the EBA stress test and the subsequent EBA capital exercise increased the capital requirements for EBA banks in 2011, the estimated 115 billion euro capital shortfall due to the capital exercise dwarfed the 2.5 billion euro capital shortfall due to the stress test (Acharya, Engle, and DianePierret, 2014). Thus, we naturally focus on the EBA capital exercise as the singularly overriding regulatory intervention.

The 61 EBA banks were selected based on total asset size. In each country, the EBA included “banks in descending order of their market shares by total assets”, such that the exercise covered “at 7 The capital exercise was an official “Recommendation” issued by the EBA. According to article 16(3) of the EBA regulation as established by the European Parliament, national supervisory authorities must make every effort to comply with the “Recommendation”. The EBA capital exercise did not coincide with other changes in the capital requirements for European banks. In particular, the EU only started with the gradual introduction of Basel III in 2013 (Capital Requirements Directive IV).

After the capital exercise, the EBA kept monitoring banks’ compliance with the 9 percent core tier 1 ratio.

least 50% of the national banking sectors in each EU Member State in terms of total consolidated assets as of end of 2010” (EBA, 2011a).8 For example, consider a country with Banks A, B, and C with 41, 30, and 10 billion euro in total assets respectively. The total size of this banking sector is 81 billion euro, with Bank A covering more than 50% of the banking sector in terms of total assets. In this example, the EBA would have included only Bank A in the exercise. As in the 2011 EBA stress test, selection into the capital exercise was based on total assets as of end of 2010 and selection was therefore not based on bank-specific events in the months prior to the capital exercise.

In contrast to the 2009 US Supervisory Capital Assessment Program (SCAP), which required banks “to raise additional capital, either in public markets or by issuing mandatory convertible preferred securities” (Hirtle, Schuermann, and Stiroh, 2009), the EBA left discretion to the banks which measures to take in order to comply with the higher capital requirements. In principle, banks can increase their capital ratios in two different ways: they can either increase their levels of regulatory capital (the numerator of the capital ratio) or they can shrink their risk-weighted assets (the denominator of the capital ratio) (Admati, DeMarzo, Hellwig, and Pfleiderer, 2010).

While raising capital is generally considered “good deleveraging” by regulators, shrinking assets has potentially adverse effects if many banks simultaneously engage in cutting lending (Hanson, Kashyap, and Stein, 2011). EBA banks increased their capital positions by more than 200 billion euro between December 2011 and June 2012. In the final report, the EBA stated that “banks’ capital strengthening has been achieved mainly via new capital measures such as retained earnings, new equity and liability management” and that “capital strengthening has not led directly to a significant reduction in lending to the real economy” (EBA, 2012).

This assessment, however, only constitutes a pure before-after comparison and does not identify EBA banks’ response to the capital exercise due to the absence of an appropriate control group. How banks adjust their balance sheets in response to the increase in capital requirements is thus an empirical question calling for a more thorough analysis.

8 From the initial 71 banks, the EBA excluded during the capital exercise banks which were “undergoing a deep restructuring”, namely Dexia, Österreichische Volksbank AG, West LB, all six Greek banks (EFG Eurobank Ergasias S.A., National Bank of Greece, Alpha Bank, Piraeus Bank Group, Agricultural Bank of Greece (ATE bank), TT Hellenic Postbank S.A.) and Bankia. We do not include these banks in the analysis. 7

II. Empirical Strategy and Data This paper exploits the 2011 EBA capital exercise to identify how banks adjust their balance sheets in response to higher capital requirements and how this adjustment process affects firms which obtain a substantial share of their borrowing from EBA banks.

Hence, we first analyze at the bank level the extent to which the exercise changed bank behavior, in particular outstanding loan volumes. Next, we move to the individual loan level in order to disentangle credit supply from credit demand. Finally, we examine the effect of higher capital requirements on asset-, investment-, and sales growth at the firm level.

A. Bank-Level Analysis The setup of the capital exercise, whereby the EBA reviewed a subset of banks’ actual capital positions and sovereign exposures and “requested them (i.e., our treatment group) to set aside additional capital buffers” (EBA, 2011b), while leaving requirements unchanged for other European banks (i.e., our control group pool), naturally lends itself to a difference-in-differences research design. However, participation in the capital exercise was not randomly assigned to banks. Instead, the EBA selected banks according to an explicit selection rule based on bank size, resulting in EBA banks being on average larger than Non-EBA banks.

This would preclude causal inference if large banks would differ from small banks, for example in terms of business models or funding strategies, and would behave differently even in the absence of a change in capital requirements. We exploit the country-specific selection threshold of the EBA selection rule to address this potential selection problem. Figure 2 shows the distribution of EBA banks and Non-EBA banks across different countries. While EBA banks are on average larger than Non-EBA banks, the country-specific selection threshold yields a considerable overlap in size between banks participating and not participating in the capital exercise.

For example, while the smallest bank included in the EBA capital exercise, the Slovenian bank Nova Kreditna banka Maribor, had 6 billion euro in total assets as of end of 2010, the largest European bank not included in the capital exercise, the French bank Crédit Mutuel, had 591 billion euro in total assets in the same year. Knowledge about the selection rule based on observable characteristics (total assets) in combination with an overlap in size allows us to combine the difference-in-differences framework with an appropriate matching 8

methodology by matching banks from the treatment group to similar banks from the control group pool. [Figure 2 about here] The paper uses the bias-corrected Abadie and Imbens (2002) matching estimator, which has recently been used by Almeida, Campello, Laranjeira, and Weisbenner (2011), Campello and Giambona (2013) and Kahle and Stulz (2013) in a corporate finance setting.9 To alleviate concerns that our results are driven by bank characteristics other than size, this paper also matches on pre-treatment levels of the core tier 1 ratio, customer loans as a share of total assets, net interest income as a share of total operating revenue, depository funding as a share of total assets, and net income over total assets.

These matching covariates capture potential differences in the capital structure, business models, funding strategies, and profitability of similarly sized banks prior to the capital exercise. The baseline matching strategy matches four Non-EBA banks to each EBA bank based on the six matching covariates.10 In addition to simply matching EBA banks to Non-EBA banks, we also employ three alternative matching strategies, each of which addresses a different potential concern. First, we match EBA banks to Non-EBA banks in the “Overlap Sample” of banks which are larger than the smallest EBA bank and smaller than the largest Non-EBA bank.

The purpose of this overlap matching strategy is to completely remove the remaining size difference between EBA banks and Non-EBA banks and to rule out that our results are driven by bank size. Second, we match EBA banks to Non-EBA banks around the selection threshold within the same country. Therefore, we construct a “Threshold Sample” which includes the two smallest EBA banks and the two largest Non-EBA banks within each country. The purpose of this within country matching strategy is to address concerns that our results are driven by cross-country differences, such as regulatory interventions and business cycles.

Finally, we use the threshold sample and match EBA banks to Non-EBA 9 In contrast to standard propensity score matching, the Abadie-Imbens estimator minimizes the Mahalanobis distance between a vector of observed matching covariates across banks in the treatment group and banks in the control group pool and introduces a bias-correction to account for inexact matches on continuous variables. The bias-corrected Abadie-Imbens matching estimator has the advantage of generally lowering the estimation bias (but increasing the variance) compared to matching estimators based on the estimated propensity score.

10 Regarding the number of matches, in our baseline specification we follow Abadie and Imbens (2011) and choose four matches, which was found to be a good trade-off between bias (which is increasing in the number of matches) and variance (decreasing in the number of matches) of the matching estimator.

banks around the selection threshold within the same region (GIIPS countries and Non-GIIPS countries). This within region matching strategy addresses the concern that our results are driven by the European sovereign debt crisis, which mainly affected banks in GIIPS countries (Acharya, Eisert, Eufinger, and Hirsch, 2016). Table I provides an overview of our baseline matching strategy and the three alternative matching strategies. [Table I about here] For all four matching strategies, we estimate the average treatment effect on the treated (ATT) using the bias-corrected Abadie and Imbens (2002) matching estimator.

The main outcome variables of interest in the bank-level part of the paper are the change in the core tier 1 ratio, the change in the logarithms of core tier 1 capital and risk-weighted assets (the components of the capital ratio), and the change in the logarithm of outstanding customer loans. For the bank-level part of the paper, we use annual bank balance sheet data from the SNL Financial Company database. Our initial sample contains all 61 EBA banks and all 494 NonEBA European commercial and savings banks from the SNL Financial universe. Since the EBA capital exercise was conducted at the highest level of consolidation, we exclude all subsidiaries of EBA banks, Non-EBA banks, and foreign banks.

As the paper wants to track the behavior of independent banks over time, we also exclude all banks which were acquired during the sample period, all banks which received capital injections during the pre-treatment period and all banks with negative levels of equity. This sample construction procedure finally leaves us with a sample of 48 EBA banks and 145 Non-EBA banks.11 The sample period spans two post-treatment years after the capital exercise (2012 and 2013) and a symmetrical time window of two pre-treatment years prior to the capital exercise (2009 and 2010).

B. Loan-Level Analysis While bank balance sheet data is appropriate for investigating how banks adjust their balance sheets in response to higher capital requirements, it is not suitable for identifying the effect on bank lending. In particular, by using bank balance sheet data one cannot disentangle credit supply 11 Table XIII lists all EBA banks in our sample. In Section III we discuss that the exclusion of 13 EBA banks does not drive our results, and that we find similar results when using the full sample of 61 EBA banks. 10

from credit demand. Thus, to study the effect of higher capital requirements on banks’ credit supply, we use loan-level data on syndicated loans and, for identification, exploit multiple bankfirm relationships in the spirit of Khwaja and Mian (2008).

As syndicated loans often have long maturities, bank exposures to individual firms are therefore often constant over time. We thus modify the estimator similar to Popov and Van Horen (2015) and Acharya, Eisert, Eufinger, and Hirsch (2016) and aggregate firms into clusters based on their industry and country of incorporation. By clustering at the country-industry level, we ensure that firms are subject to the same regional and sectoral shocks over time and attribute the remaining variation in loan exposure volumes to credit supply shocks.

  • We then estimate the following difference-in-differences regression specification: ΔLoan Exposurebij = β · EBA Bankbi + γ · Xbi + ηi + ηj +
  • bij (1) where ΔLoan Exposurebij is the change in loan exposure of bank b in country i to firm cluster j between the five quarters before the EBA capital exercise (2010Q2 - 2011Q2) and the five quarters after the capital exercise (2012Q3 - 2013Q3). The variable EBA Bankbi takes on the value of 1 if the bank is part of the EBA capital exercise, and 0 otherwise. In addition, the specification includes bank characteristics as of 2010 (log total assets, core tier 1 ratio, customer loans as a share of total assets, net interest income as a share of total operating revenue, depository funding as a share of total assets, and net income over total assets) and firm cluster fixed effects ηj, which absorb all cluster-specific credit demand shocks. Moreover, we include bank country fixed effects ηi to absorb country-specific shocks, which affect all banks in a given country. Like Khwaja and Mian (2008), we follow Bertrand, Duflo, and Mullainathan (2004) and collapse our data into a preand post-treatment period before differencing in order to produce standard errors that are robust to concern of autocorrelation. In addition, standard errors are clustered at the bank level. For the loan-level part of the paper, we obtain data from Thomson Reuters LPCs Dealscan database, which contains detailed information on syndicated loan contract terms, loan types, and maturities. We collect data on all outstanding term loans and credit lines from banks in our sample to non-financial corporate borrowers incorporated in EBA countries.12 Of the 76 banks 12 For term loans and credit lines, we follow the variable definition of Berg, Saunders, and Steffen (2016). 11
  • in our matched control group, 63 were active in the syndicated loan market during our sample period and are feasible to serve as control group banks in the loan-level part of the paper. Dealscan contains full information on the loan allocation between syndicated members for about 32% of all loans. For the remaining 68%, we follow De Haas and Van Horen (2012) and divide the loan facility equally among all members of a syndicate. Our initial sample contains 10,829 syndicated loans from 109 banks to 5,693 companies. The LPC Dealscan database contains the issuance of new syndicated loans at the time of origination. In order to employ our modified version of the Khwaja and Mian (2008) estimator, we transform the data and calculate the outstanding exposure of bank b in country i to firm cluster j in quarter q using the maturity variable contained in the database. In our main analysis, we focus on the intensive margin sample which includes only firm clusters to which EBA banks lend both before and after the capital exercise. Thus, this sample excludes firm clusters that entirely stop borrowing after the EBA capital exercise or do not borrow prior to the capital exercise. The intensive margin sample includes 45 EBA banks and 44 Non-EBA banks. In Section IV, we provide additional results on the extensive margin sample of firms. C. Firm-Level Analysis In the final empirical step, we link the EBA banks’ balance sheet adjustments to real outcomes at the firm level. A reduction in credit supply of EBA banks would not necessarily yield effects at the firm level if other banks, not subject to an increase in capital requirements, would pick up the slack. An increase in capital requirements for the subset of EBA banks would then not affect the total supply of credit to the real economy and would not affect firms’ corporate policies. In order to measure a firm j’s dependence on credit supply from EBA banks prior to the capital exercise, we construct the variable EBA Borrowing Sharej: EBA Borrowing Sharej =
  • i[EBABanks] 1 5
  • 2011Q2 q=2010Q2 OutstandingLoansijq
  • i[AllBanks] 1 5
  • 2011Q2 q=2010Q2 OutstandingLoansijq (2) where the numerator is the average amount of outstanding loans of firm j obtained from EBA banks over the five quarters prior to the capital exercise (2010Q2 - 2011Q2) and the denominator is the average amount of total outstanding loans of firm j obtained from all banks over the same period. For firms in our sample which were not borrowing in the syndicated loan market in the 12

period before the capital exercise (but in the period after the capital exercise), we assign a EBA borrowing share of zero, since those firms were not dependent on credit supply from EBA banks prior to the capital exercise. In the bankand loan level part, we restrict our analysis to banks from EBA countries. Since European firms might also borrow from banks incorporated in Non-European countries, we now also include those banks when computing the EBA borrowing share. We then divide our sample of firms into “EBA firms” with an above median dependence on credit supply from EBA banks as measured by the EBA borrowing share (our treatment group), and “Non-EBA firms” with a below median dependence on credit supply from EBA banks (our control group pool).

Since EBA firms might differ from Non-EBA firms along a number of important characteristics, we employ a difference-in-differences matching methodology analog to the one used in the bank-level part. In the firm-level part of the paper, we match firms on country of incorporation, industry as defined by the 1-digit SIC code, whether the firm is publicly listed or not, and pre-treatment levels of the logarithm of total assets, tangibility, cash flow over total assets, net worth, EBITDA over total assets, and leverage.13 As in the bank-level part of the paper, we estimate the treatment effect on the treated (ATT) using the Abadie and Imbens (2002) bias-corrected matching estimator.

The main outcome variables in the firm-level part of the paper are the changes in the logarithm of total assets, fixed assets (as a measure of investment, following Campello and Larrain (2016)), and sales between the period before the capital exercise (2009 and 2010) and after the capital exercise (2012 and 2013). All variables are winsorized at the 5% level.14 As we expect results to be stronger for firms which are less likely to substitute a reduction in credit supply with other sources of funding (e.g. issuing equity), we also split our sample into listed and unlisted firms and report results separately.

For the firm-level part of the paper, we use information on firms’ balance sheets and profit and loss statements from Bureau van Dijk’s Amadeus Financials database. The database additionally contains information on a firm’s country of incorporation, its SIC industry code, and whether the firm is publicly listed. We have access to the sample of firms classified as Very Large, Large, and Medium by Amadeus. Since the Dealscan database and the Amadeus database share no common identifier, we hand-merge the two datasets and additionally require non-missing values on 13 The definitions of all variables are summarized in Table XIV.

14 Regarding the level of winsorization we follow Acharya, Eisert, Eufinger, and Hirsch (2016). In unreported robustness tests, we find similar results when winsorizing the variables at the 1% level. 13

all relevant variables, which leaves us with a sample of 1,958 firms for the firm-level part of the paper. III. Results In this section, we present the empirical results for the bank-level-, loan-level-, and firm-level part of the paper. A. Bank-Level Results We first provide summary statistics before and after matching under the different matching strategies. Table II shows the median pre-treatment values of the matching covariates for EBA banks, Non-EBA banks, and control group banks as of end of 2010, the year immediately prior to the capital exercise. The paper uses the continuity corrected Pearson χ2 test statistic to test for differences in medians between the groups.

[Table II about here] Panel A of Table II compares the 48 EBA banks with 145 Non-EBA banks. As expected, EBA banks significantly differ from Non-EBA banks along a number of important dimensions. Due to the capital exercise being carried out on the largest banks in each country, the median EBA bank is more than 20 times larger than the median Non-EBA bank. The two groups of banks also significantly differ in terms of their business models, with the median EBA bank being less engaged in customer lending and generating less of its revenue from interest income than the median Non-EBA bank. While the two groups of banks do not differ significantly with regard to their pre-treatment core tier 1 ratios, the median EBA bank is significantly less reliant on customer deposits (i.e.

more reliant on wholesale funding) than the median Non-EBA bank. These large differences between EBA banks and Non-EBA banks regarding important characteristics emphasize the necessity of employing a matching procedure.

For our baseline matching strategy, we match four Non-EBA banks to each EBA bank based on the Mahalanobis distance of all matching covariates as of end of 2010. Panel B of Table II shows the median values of EBA banks and control groups banks based on our baseline matching 14

specification. The matching procedure significantly reduces the differences between EBA banks and Non-EBA banks, especially with regard to our measures for banks’ business models. Some differences, however, remain significant, albeit at a much lower level compared to the unmatched sample.

While EBA banks are still bigger than control group banks, our matching procedure reduces the difference from EBA banks being more than 20 times bigger to EBA banks being roughly 10 times bigger. To address concerns that our results might be driven by bank size, we employ the overlap matching strategy, which includes all banks larger than the smallest EBA bank and smaller than the largest Non-EBA bank, and match one Non-EBA bank to each EBA bank based on asset size only.15 Panel C of Table II shows that this matching strategy reduces the size difference to 30 billion euro, which is statistically insignificant.

A further concern might be that our results are driven by country-specific factors, such as differences in macroeconomic environments or regulatory interventions. To rule this out, we employ the within country matching strategy in Panel D, which matches the two largest Non-EBA banks to the two smallest EBA banks within each country using the threshold sample. To specifically address concern that our results are driven by banks from GIIPS countries, which were especially exposed to the European sovereign debt crisis, we employ the within region matching strategy in Panel E, which uses the threshold sample and matches EBA banks to Non-EBA banks around the selection threshold within the same region (GIIPS countries and Non-GIIPS countries).16 We first examine whether EBA banks did indeed increase their core tier 1 ratios in response to higher capital requirements, and whether they did so via increasing their levels of capital (adjustment via the numerator) or via reducing risk-weighted assets (adjustment via the denominator).

Identification in a difference-in-differences framework crucially relies on the parallel trend assumption to hold. We therefore examine the pre-treatment trend in the core tier 1 ratio for both EBA banks and the sample of matched control group banks according to our baseline matching strategy. Figure 3 shows a general upward trend in the core tier 1 ratios for both groups. This increase is parallel up to 2010, the year immediately prior to the capital exercise. Starting in 2011, EBA banks began to increase their core tier 1 ratios significantly more than banks in the matched control 15 In addition, we employ a difference-in-differences regression methodology in Section IV to test whether our results are driven by above median size banks.

16 To test whether the results are driven by unobserved country-specific time-varying characteristics, the differencein-differences regression framework in Section IV includes country-year fixed effects. 15

group. [Figure 3 about here] Figure 4 and Figure 5 show the trends in core tier 1 capital (the numerator of the core tier 1 ratio) and risk-weighted assets (the denominator of the core tier 1 ratio) respectively, normalized to the value of 1 for the year 2010. Both the levels of core tier 1 capital and risk-weighted assets evolved parallel for EBA banks and matched control group banks in the years leading up to the capital exercise.

However, while levels of core tier 1 capital continued their parallel increase after the capital exercise, EBA banks started reducing their risk-weighted assets significantly compared to banks in the matched control group.

[Figure 4 and Figure 5 about here] Figures 3 to 5 provide evidence that our baseline matching procedure does a good job at balancing EBA banks and Non-EBA banks with regard to their pre-treatment trends in the core tier 1 ratio and its components. Moreover, the graphs indicate that EBA banks increased their core tier 1 ratios compared to the matched control group and that they did so, not by increasing their levels of core tier 1 capital, but by reducing their levels of risk-weighted assets. For the different matching strategies, we estimate the differences in the change in the core tier 1 ratio, the logarithm of core tier 1 capital and the logarithm of risk-weighted assets from the period before to the period after the capital exercise between EBA banks and banks in the matched control group.

The first column of Panel A of Table III shows how both EBA banks and banks in the matched control group adjusted their core tier 1 ratios around the 2011 EBA capital exercise. Row 1 reports the before-after differences for EBA banks, Row 2 the before-after differences for control group banks, Row 3 the unmatched difference-in-differences results, and Row 4 the bias-corrected Abadie and Imbens (2002) matching estimator for the average treatment effect on the treated (ATT). Both EBA banks and control group banks increased their core tier 1 ratios in the two years after the capital exercise, reflecting a general upward trend among European banks, which can also be seen in Figure 3.

However, while control group banks increased their core tier 1 ratios by only 1.79 percentage points on average, EBA banks did so by 3.02 percentage points and thus significantly more than banks not subject to higher capital requirements. The ATT equals 16

1.85 percentage points and is significant at the 1% level, indicating that the increase in capital requirements did indeed affect the core tier 1 ratios of banks participating in the capital exercise. [Table III about here] The second column of Panel A of Table III shows that EBA banks increased their levels of core tier 1 capital by 19% around the 2011 EBA capital exercise. However, as the comparison with the matched control group indicates, this increase seems to reflect a general development in the European banking system, e.g. market pressure on banks to raise capital, rather than an effect of the capital exercise.

European banks not participating in the capital exercise exhibited an almost identical percentage increase in their levels of core tier 1 capital, rendering the ATT insignificant. This finding provides evidence that EBA banks did not respond to the increase in capital requirements by raising new capital. In contrast, there is a significant difference in the change of risk-weighted assets between EBA banks and matched control group banks around the capital exercise, as can be seen in the third column of Panel A of Table III. While EBA banks reduced their levels of risk-weighted assets by 10% over our sample period, control group banks even increased their levels of risk-weighted assets.

The ATT indicates that EBA banks reduced their risk-weighted assets by 16 percentage points compared to banks in the matched control group which were not subject to an increase in capital requirements. The combined findings in Table III are the first central result of the bank-level part of the paper. They provide evidence that banks, when faced with an increase in capital requirements, adjust their capital ratios by reducing their levels of risk-weighted assets (adjustment via the denominator) rather than by raising new capital (adjustment via the numerator).17 The analog matching results of the overlap matching strategy in Panel B, the within country matching strategy in Panel C, and the within region matching strategy in Panel D of Table III show that our results are robust to concerns of bank size, country-specific factors, and exposure to the European sovereign debt crisis respectively.

In all cases, the matching results suggest that EBA banks responded to the increase in capital requirements by reducing their risk-weighted assets compared to banks in the control group.

17 When using the sample of all 61 EBA banks, we find that EBA banks increased their core tier 1 ratio by 2.13 percentage points and reduced their risk-weighted assets by 15 percentage points compared to banks in the matched control group. Both estimates are significant at the 1% level. Furthermore, we find no significant difference with regard to the levels of core tier 1 capital. 17

EBA banks could achieve this relative reduction in risk-weighted assets either by changing their asset composition from riskier assets into safer assets while keeping total asset size constant (risk reduction), or by reducing total asset size while keeping the average asset risk constant (asset shrinking).

Asset shrinking has potential negative effects on the real economy if a large fraction of banks simultaneously decides to reduce lending. In order to examine whether EBA banks reduced their risk-weighted assets via risk reduction, Panel A of Table IV reports the matching estimation results for two different measures of banks’ asset risk as the outcome variable. Risk reduction behavior would imply a reduction in the ratio of risk-weighted assets over total assets. The first column, however, shows that there is no statistically significant difference in the changes of this ratio between EBA banks and banks in the matched control group.

While EBA banks did indeed reduce risk-weighted assets relative to total assets (see also Acharya and Steffen (2015)), so did banks in the matched control group, again reflecting a general development in the European banking system rather than an effect of higher capital requirements. Similarly, the second column shows that there is neither a significant treatment effect with regard to loan loss reserves relative to outstanding customer loans.

[Table IV about here] In Panel B of Table IV, we investigate whether EBA banks reduced their risk-weighted assets by engaging in asset shrinking. The first column shows that EBA banks reduced total assets by 14 percentage points compared to banks in the matched control group. The EBA’s final report on the capital exercise claims that “capital plans have not led directly to a significant reduction of lending into the real economy” (EBA, 2012). The second column of Panel B shows that EBA banks did indeed not reduce outstanding customer loans after the capital exercise. However, control group banks increased customer loans by 8 percentage points during the same period.

The matching estimator indicates that EBA banks reduced outstanding customer loans by 12 percentage points compared to the matched control group of banks not subject to an increase in capital requirements. We also document a negative treatment effect on security holdings of EBA banks.18 However, as customer loans make up 60% of the average EBA bank’s balance sheet while security holdings only 18 Boyson, Helwege, and Jindra (2014) find that when banks “sell assets, they cherry pick assets in order to alleviate pressure from capital regulations”.

You can also read
Next part ... Cancel